Artificial Neural Networks Models for Identifying Flow Regimes and Predicting Liquid Holdup in Horizontal Multiphase Flow

Abstract

Abstract Knowledge of liquid holdup and multiphase flow regimes in pipes is very important in pressure drop calculations and design of surface facilities. To predict the holdup, one has to define the flow regime at a given pressure, superficial gas and liquid velocities, temperature, and other flow characteristics. Several flow maps, regression and mechanistic models are available in literature to predict holdup and flow regime. However, some of these models do not predict the liquid holdup correctly. This paper presents two Artificial Neural Networks (ANN) models to identify the flow regime and calculate the liquid holdup in horizontal multiphase flow. These models are developed using experimental data - 199 data points - and utilizing three-layer back propagation neural networks. Superficial gas and liquid velocities, pressure, temperature and fluid properties are used as inputs to the network. The output of the first network is the flow regime, while the other network predicts the liquid holdup. One-half of the data was used to train the ANN models, one quarter to cross-validate the relationships established during the training process and the remaining quarter to test the models and evaluate their accuracy. The results show that the developed models provide better predictions and higher accuracy than the empirical correlations developed specifically for these data groups. The developed flow regime model predicts the flow regime correctly for more than 97% of the data points. The liquid holdup model outperforms the published models in terms of the lowest absolute average percent error (9.407), the lowest standard deviation (8.544) and the highest correlation coefficient (0.9896).